K Schwab1, M Eiselt, C Schelenz, H Witte. 1. Institute of Medical Statistics, Computer Science and Documentation, Medical Faculty of the Friedrich Schiller University of Jena, Jena, Germany. iks@imsid.uni-jena.de
Abstract
OBJECTIVES: Electroencephalographic burst activity characteristic of burst-suppression pattern (BSP) in sedated patients and of burst-interburst pattern (BIP) in the quiet sleep of healthy neonates have similar linear and non-linear signal properties. Strong interrelations between a slow frequency component and rhythmic, spindle-like activities with higher frequencies have been identified in previous studies. Time-varying characteristics of BSP and BIP prevent a definite patternrelated analysis. A continuous estimation of the bispectrum is essential to analyze these patterns. Parametric bispectral approaches provide this opportunity. METHODS: The adaptation of an AR model leads to a parametric bispectrum by using the transfer function of the estimated AR filter. Time-variant parametric bispectral approaches require an estimation of AR parameters which consider higher order moments to preserve phase information. Accordingly, a time-variant parametric estimation of the bispectrum was introduced. Data driven simulations were performed to provide optimal parameters. BSP (12 patients) and BIP (6 neonates) were analyzed using this novel approach. RESULTS: Significant differences in the time course of burst pattern during BSP and burst-like pattern before the onset of BSP could be shown. A rhythmic quadratic phase coupling (period 10 sec) was identified during BIP in all neonates. CONCLUSION: Quadratic phase couplings during BSP increases in the time course depending on depth of sedation. The visually detected burst activity in BIP is only the temporarily observable EEG correlate of a hidden neural process. Time-variant bispectral approaches offer the possibility of a better characterization of underlying neural processes leading to improved diagnostic tools used in clinical routine.
OBJECTIVES: Electroencephalographic burst activity characteristic of burst-suppression pattern (BSP) in sedated patients and of burst-interburst pattern (BIP) in the quiet sleep of healthy neonates have similar linear and non-linear signal properties. Strong interrelations between a slow frequency component and rhythmic, spindle-like activities with higher frequencies have been identified in previous studies. Time-varying characteristics of BSP and BIP prevent a definite patternrelated analysis. A continuous estimation of the bispectrum is essential to analyze these patterns. Parametric bispectral approaches provide this opportunity. METHODS: The adaptation of an AR model leads to a parametric bispectrum by using the transfer function of the estimated AR filter. Time-variant parametric bispectral approaches require an estimation of AR parameters which consider higher order moments to preserve phase information. Accordingly, a time-variant parametric estimation of the bispectrum was introduced. Data driven simulations were performed to provide optimal parameters. BSP (12 patients) and BIP (6 neonates) were analyzed using this novel approach. RESULTS: Significant differences in the time course of burst pattern during BSP and burst-like pattern before the onset of BSP could be shown. A rhythmic quadratic phase coupling (period 10 sec) was identified during BIP in all neonates. CONCLUSION: Quadratic phase couplings during BSP increases in the time course depending on depth of sedation. The visually detected burst activity in BIP is only the temporarily observable EEG correlate of a hidden neural process. Time-variant bispectral approaches offer the possibility of a better characterization of underlying neural processes leading to improved diagnostic tools used in clinical routine.